From owner-freebsd-isp Sun Dec 16 15:59:41 2001 Delivered-To: freebsd-isp@freebsd.org Received: from lemon.national.com.au (lemon.national.com.au [203.57.241.81]) by hub.freebsd.org (Postfix) with ESMTP id E456C37B41A for ; Sun, 16 Dec 2001 15:59:37 -0800 (PST) Received: by lemon.national.com.au (Postfix, from userid 5) id 586399F821; Mon, 17 Dec 2001 10:59:31 +1100 (EST) Received: from unknown(10.25.154.32) by lemon.national.com.au via csmap (V4.1) id srcAAAlNaisS; Mon, 17 Dec 01 10:59:30 +1100 Received: (from uucp@localhost) by peppermint.national.com.au (8.9.3+Sun/8.8.8) id KAA15191; Mon, 17 Dec 2001 10:59:23 +1100 (EST) Received: from webjump.national.com.au(164.53.27.38) via SMTP by peppermint, id smtpdAAAQZaOQD; Mon Dec 17 10:59:23 2001 Received: (from nconedd@localhost) by webjump.national.com.au (8.10.2+Sun/8.10.2) id fBGNxQS22427; Mon, 17 Dec 2001 10:59:26 +1100 (EST) Date: Mon, 17 Dec 2001 10:59:26 +1100 From: Enno Davids To: Colin Campbell Cc: Jeff Lasman , freebsd-isp@FreeBSD.ORG Subject: Re: Using DNAT and DNS round-robin Message-ID: <20011217105926.K16592@webjump.national.com.au> References: <3C1D0EF1.783B48AD@nobaloney.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2.5i In-Reply-To: ; from sgcccdc@citec.qld.gov.au on Mon, Dec 17, 2001 at 09:49:19AM +1000 Sender: owner-freebsd-isp@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.org On Mon, Dec 17, 2001 at 09:49:19AM +1000, Colin Campbell wrote: |On Sun, 16 Dec 2001, Jeff Lasman wrote: |> Derrick John Klise wrote: |> |> > IIRC, something like: |> > |> > monkey.example.net IN A 192.168.0.1 |> > IN A 192.168.0.2 |> > IN A 192.168.0.3 |> |> Thanks. Finally found it on page 259 of DNS and Bind. |> |> > > Is there a way to handle high-availability strictly in DNS? |> > | |There used to be (still is? - cou;dn't find it) a paper on the ISC web |site (www.isc.org) exlpaining why using DNS for HA was pointless. If |memory serves, the main reasons were | |- most browsers cache DNS lookups and so a system that goes down will | simply appear as unreachable to the browser. | |- most browsers ignore TTLs. FWIW, squid (and possibly other proxies) when faced with a list of address for a name will retry on the next address in the list when they get a hard error on the one they're using. Its still not HA, but its better than you thought. The real answer is hardware load balancers like F5, Foundry or Rad. Enno. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-isp" in the body of the message